Reinforcement Learning-Based Framework for Structural Design Optimization

Min-Jeong Cho; Hyeon-Jong Hwang

doi:10.23310/PF.2026.3.2.089

All Issue

2026 Vol.3, Issue 2 Preview Page Next Page

Research Article

Reinforcement Learning-Based Framework for Structural Design Optimization 구조물 최적 설계를 위한 강화학습 기반 방법

30 April 2026. pp. 89-99

PDF XML

Abstract

This study proposes a universal structural design framework based on Reinforcement Learning (RL) to maximize structural safety and to optimize material efficiency. Conventional numerical optimization techniques often face challenges, such as excessive computational demands or convergence to local optima in large-scale non-linear design spaces. To address these issues, a customized environment integrating OpenAI Gymnasium with a structural analysis engine was developed, employing the Proximal Policy Optimization (PPO) algorithm. The effectiveness of the proposed framework was validated through the sectional optimization of concrete-filled U-shaped steel composite beams and truss structures. The results demonstrated that the RL agent successfully learned the complex non-linear relationships between sectional geometry and structural performance, effectively aligning the structural capacity with the target requirements while achieving superior material efficiency, compared to conventional sections. This research confirms that RL can autonomously establish optimal design policies that strictly satisfy design code requirements, suggesting significant potential as a decision-making tool for automated structural design and optimization in practice.

Keywords

강화학습

구조 최적화

PPO 알고리즘

합성보

트러스

본 연구에서는 구조물의 안전성 극대화와 재료 효율성 최적화를 위한 강화학습 기반의 범용 구조설계 프레임워크를 제안한다. 기존의 수치 최적화 기법은 설계 변수가 방대하고 제약 조건이 비선형적인 환경에서 과도한 계산 시간이 소요되거나 국부 최적해에 수렴하는 한계가 있다. 이를 극복하기 위해 본 연구에서는 OpenAI Gymnasium과 구조 해석 엔진이 통합된 사용자 정의 환경을 구축하고, PPO 알고리즘을 적용하였다. 제안된 프레임워크의 유효성을 검증하기 위해 충전형 U형 강재 합성보와 트러스 구조의 단면 최적화를 수행하였다. 학습 결과, 강화학습 에이전트는 단면 치수와 구조 성능 사이의 복잡한 비선형 관계를 효과적으로 학습하여 부재 성능을 요구 성능에 최적으로 근접시켰으며, 기성 제품 대비 향상된 재료 효율성을 달성하였다. 본 연구는 강화학습이 설계 요구조건을 만족하는 최적 설계 정책을 스스로 수립할 수 있음을 입증하였으며, 향후 구조설계 자동화 및 최적화를 위한 의사결정 도구로 활용 가능함을 보였다.

키워드

Reinforcement Learning

Structural Optimization

Proximal Policy Optimization (PPO)

Composite Beam

Truss

References

Kim, J. C., Kim, D. K., Kim, J. H., Jung, J. W., Lee, S. H., and Lee, C. N. (2018) Design Method According to Eurocode for TSC Composite Beams Applied to Singapore. Magazine of the Korean Society of Steel Construction 30(3), 18-25. (In Korean)

Lemus-Romani, J., Ossandón, D., Sepúlveda, R., Carrasco-Astudillo, N., Yepes, V., and García, J. (2023) Optimizing Retaining Walls Through Reinforcement Learning Approaches and Metaheuristic Techniques. Mathematics 11(9), 1-33.

10.3390/math11092104

Ministry of Land, Infrastructure and Transport (MOLIT) (2022) Design Standard for Building Steel-Concrete Composite Structures (KDS 41 30 20:2022). Seoul: Architectural Institute of Korea. (In Korean)

Raffin, A., Hill, A., Gleave, A., Kanervisto, A., Ernestus, M., and Dormann, N. (2021) Stable-Baselines3: Reliable Reinforcement Learning Implementations. Journal of Machine Learning Research 22(268), 1-8.

Rochefort-Beaudoin, T., Vadean, A., Aage, N., and Achiche, S. (2025) Structural Design Through Reinforcement Learning. Engineering Applications of Artificial Intelligence 161, 112273.

10.1016/j.engappai.2025.112273

Schulman, J., Levine, S., Moritz, P., Jordan, M. I., and Abbeel, P. (2015) Trust Region Policy Optimization. In Proceedings of the 32nd International Conference on Machine Learning, Lille 7-9 July 2015. PMLR. 1889-1897.

Schulman, J., Wolski, F., Dhariwal, P., Radford, A., and Klimov, O. (2017) Proximal Policy Optimization Algorithms. arXiv preprint arXiv:1707.06347.

Sutton, R. S. and Barto, A.G. (2018) Reinforcement Learning: An Introduction. 2nd edtion. Cambridge, MA: MIT Press.

Towers, M., Kwiatkowski, A., Terry, J., Balis, J. U., De Cola, G., Deleu, T., Goulão, M., Kallinteris, A., Krimmel, M., KG, A., Perez-Vicente, R., Pierré, A., Schulhoff, S., Tai, J. J., Tan, H., and Younis, O. G. (2024) Gymnasium: A Standard Interface for Reinforcement Learning Environments. arXiv preprint arXiv:2407.17032.

Zhao, G., Ma, G., Cho, M. J., and Hwang, H. J. (2026) Automated Design of Shear Wall Structures Using Reinforcement Learning with Code Compliance. Engineering Structures 353, 122272.

10.1016/j.engstruct.2026.122272

Information

Publisher :Korea Protective Facility Institute
Publisher(Ko) :한국방호시설학회
Journal Title :Protective Facility
Journal Title(Ko) :방호시설
Volume : 3
No :2
Pages :89-99
Received Date : 2026-03-23
Revised Date : 2026-04-05
Accepted Date : 2026-04-07
DOI :https://doi.org/10.23310/PF.2026.3.2.089

Protective Facility ISSN:3058-7816(Print) 3058-8618(Online) 방호시설

All Issue